Why is latency/ping at such odds with other tests?

[jackdashack]

Every test I do .. about 6 others consistently rate me in the 16-18ms ping. Why is my ping here always so high? 59 to 65. I know this is a non-flash with better results but Mytest has never been that much at odds. One is as high as 500+. I consistently get the low above across other speed test. Of course I am very interested in ping and then speed. Thanks so much! jack ":-Dx

[CA3LE]

There's a reason I don't call it latency or ping.  I'm also sorry for not providing any documentation but the truth is... I'm still trying to understand that tool myself.  Not finished.

 

What it's testing is the time it takes for you to respond to a very simple request, the round trip time is then cut in half.  This number can be very similar to the number you'd get from ping, +1-5ms when things are totally ideal.  It can be influenced by the device, browser, software installed on top of the browser and even open tabs.  Keep in mind that if your result is affected in the test, chances are that the other browser tabs your surfing are also experiencing the same lag.  If you see a consistent 500+ms RT something is causing it to happen. If it spikes and you get that high reading more than one time in a row it's usually a bad sign.

 

I often leave tons of tabs open.  Sometimes it can get to where Chrome, Safari and Firefox all have 20-40 tabs open at the same time.  Rarely I see my RT (Response Time) spike up on TMN... look.

 

 

The few seesaw spikes don't concern me, it's the occurrences where it goes over 200ms and stayed there.  Each time that happened I closed all the browser tabs and restarted the browser... immediately RT returned to baseline.  The test didn't just do that for no reason... something influenced it and if it can influence my test it can influence the other tabs your browsing on.  Whatever the cause, closing the open tabs and restarting the browser cleared it up... meaning there had to be a true issue.  Maybe a script or something running on one of those tabs causing increased latency.  Opening another browser showed normal RT (you can see on the instance 3/4 in where the second result dropped back down, that's when I opened the other browser to compare) and ping from terminal to the selected server was also normal.

 

Keep in mind that ping influences RT but RT doesn't always affect ping because it can be influenced by things that don't affect ping.  Originally I was building a latency/ping test.  I built it the way I thought would work and on accident ended up building something different, my observations told me that I couldn't really call it ping so I call it Response Time.  

 

Unless you see it go way off baseline and stay there I'd be less concerned about little bobbles and be looking more at the average.  This isn't perfect, nothing here is or will ever be... but it's helped me detect issues more than a few times so I released it to our members hoping it will help more people.  

[jackdashack]

36 minutes ago, CA3LE said:

There's a reason I don't call it latency or ping.  I'm also sorry for not providing any documentation but the truth is... I'm still trying to understand that tool myself.  Not finished.

 

What it's testing is the time it takes for you to respond to a very simple request, the round trip time is then cut in half.  This number can be very similar to the number you'd get from ping, +1-5ms when things are totally ideal.  It can be influenced by the device, browser, software installed on top of the browser and even open tabs.  Keep in mind that if your result is affected in the test, chances are that the other browser tabs your surfing are also experiencing the same lag.  If you see a consistent 500+ms RT something is causing it to happen. If it spikes and you get that high reading more than one time in a row it's usually a bad sign.

 

I often leave tons of tabs open.  Sometimes it can get to where Chrome, Safari and Firefox all have 20-40 tabs open at the same time.  Rarely I see my RT (Response Time) spike up on TMN... look.

 

... something influenced it and if it can influence my test it can influence the other tabs your browsing on.  Whatever the cause, closing the open tabs and restarting the browser cleared it up... meaning there had to be a true issue.  Maybe a script or something running on one of those tabs causing increased latency.  Opening another browser showed normal RT (you can see on the instance 3/4 in where the second result dropped back down, that's when I opened the other browser to compare) and ping from terminal to the selected server was also normal.

 

Understood, good info. However when I run these test it's usually back to back, in sequence, same issues involved. And I've done this running of sequence of tests often and get the low PING say on SpeedTest.net (Yeah, I know your stomach just turned .. )  But I run about 5 like that and TestMy always gives a 2x or more higher ping or whatever the "ms" stat means. Should I cut that in half? Still cut in half not nearly as good as the other multiple stats. When techs come from Verizon or Comcast they all say they don't worry so much about the speed ... it's the ping they look at. ??? No worry, Just curious ... no need to get hung up on this long.   Others have hard problems here in the I suspect in the forum. It's been a long long time since I've visited. Nice to see CA3LE you are still in there and doing well. You've helped me well in the past.  Jack  ":-Dx

 

[CA3LE]

1 hour ago, jackdashack said:

And I've done this running of sequence of tests often and get the low PING say on SpeedTest.net (Yeah, I know your stomach just turned .. )  But I run about 5 like that and TestMy always gives a 2x or more higher ping or whatever the "ms" stat means. Should I cut that in half?

 

No you shouldn't cut it in half.  It's already been cut in half by the time you see it.  Your ping to speedtest may also be on a server that's physically closer than the ones you're using here.

 

To do this open your command prompt (looks to me like you're on XP).  Type

ping east.testmy.net

that's for the Washington D.C. server you're using, if you're testing against another server substitute it's subdomain...

 

Central US >> Dallas, TX, USA - dallas

Central US >> Colorado Springs, CO, USA - co

East Coast US >> Washington, D.C. - east

East Coast US >> New York, NY, USA - ny

West Coast US >> San Jose, CA, USA - west

West Coast US >> San Francisco, CA, USA - sf

Europe >> London, GB - uk

Europe >> Dusseldorf, DE - de

Asia >> Hong Kong, CN - cn

Asia >> Singapore, SG - sg

Australia >> Melbourne, AU - au

 

Let me give you my numbers as an example.  Ethernet wired connection, Comcast Extreme 150.  First I'll show you the actual ping in terminal, then TMN RT... I did an average of 10 pings.  e.g.

ping -c 10 dallas.testmy.net

 

ping -c 10

 

Central US >> Dallas, TX, USA - round-trip min/avg/max/stddev = 25.587/26.383/28.471/0.874 ms

Central US >> Colorado Springs, CO, USA - round-trip min/avg/max/stddev = 13.817/14.565/16.175/0.660 ms

East Coast US >> Washington, D.C. - round-trip min/avg/max/stddev = 57.212/57.706/58.733/0.463 ms

East Coast US >> New York, NY, USA - round-trip min/avg/max/stddev = 58.659/59.323/59.995/0.385 ms

West Coast US >> San Jose, CA, USA - round-trip min/avg/max/stddev = 39.326/39.952/40.580/0.400 ms

West Coast US >> San Francisco, CA, USA - round-trip min/avg/max/stddev = 53.522/54.429/58.288/1.474 ms

Europe >> London, GB - round-trip min/avg/max/stddev = 151.804/160.726/170.455/7.082 ms

Europe >> Dusseldorf, DE - round-trip min/avg/max/stddev = 144.740/146.039/153.047/2.358 ms

Asia >> Hong Kong, CN - round-trip min/avg/max/stddev = 204.042/204.488/205.078/0.319 ms

Asia >> Singapore, SG - round-trip min/avg/max/stddev = 211.519/213.094/213.930/0.744 ms

Australia >> Melbourne, AU - round-trip min/avg/max/stddev = 309.783/310.274/311.033/0.355 ms

 

The Response Time by TestMy.net

 

Central US >> Dallas, TX, USA - 37 ms / 35 ms avg

Central US >> Colorado Springs, CO, USA - 22 ms / 20 ms avg

East Coast US >> Washington, D.C. - 64 ms / 67 ms avg

East Coast US >> New York, NY, USA - 64 ms / 70 ms avg

West Coast US >> San Jose, CA, USA - 44 ms / 48 ms avg

West Coast US >> San Francisco, CA, USA - 57 ms / 62 ms avg

Europe >> London, GB - 160 ms / 170 ms avg

Europe >> Dusseldorf, DE - 152 ms / 154 ms avg

Asia >> Hong Kong, CN - 237 ms / 228 ms avg

Asia >> Singapore, SG - 211 ms / 217 ms avg

Australia >> Melbourne, AU - 323 ms / 328 ms avg

 

Very similar.  But like I said, I've seen things influence RT and not ping.  You should compare your ping to your Response Time but you can't really compare your Response Time to your ping.  Response Time is always slightly higher than the actual ping because of request headers and other factors involved.  I don't correct for anything, you're seeing the raw information as the program sees it... "it took X amount of time for your browser to receive and respond to my super tiny request."  ...then the time is cut in half to represent one way.

[jackdashack]

Yes, DC is my default and never Dallas. lol  And I think I follow your reasoning which I know is correct. So ... I'll read it several times until it sinks. Thanks again so very much! 

[jackdashack]

I just ran 2 tests here and got below 1mg download. Quickly did Megapath and got 2.7mg download  and 17ms latency. So not sure what's going on but no worries. It seems fine. BTW brother was just up from FL and we ran a dedicated DSL shielded line from box to primary Verizon router. We also ran a dedicated shielded Ethernet line to the slave router in the kitchen. And yes, this kitchen one is an XP but the one at the primary is a Windows 10. Just FYI

[CA3LE]

Well something happened...

 

Your speed dropped and RT skyrocketed.  I see that before it happened the identifier was 'kitchen' is this a different machine?

 

Your RT is as low as 29ms sometimes so it's not far off from what your seeing on the other test.  The question is why doesn't it stay there.  Look at mine to the server in Colorado Springs https://testmy.net/rt/CA3LE?miq=18&isp=comcast_cable.

 

A few spikes but it's fairly consistent with a low average, an average that's close to the lowest time.  That's good.

 

Your average on the other hand is much higher than your lowest Response Time.  That's worse but it doesn't necessarily mean that your connection has issues.  If you see it happen on more than one machine you should open command prompt and ping the test server and see if the issue is affecting the actual ping.

 

Regardless of what the other tests tell you, you shouldn't ignore the results here.  For all we know the other tests may run a series of pings and only keep the best one.

 

The RT numbers don't just come out of the air.  If everything between you and the test server and all other variables were perfect you'd see a reading of 0ms... my local testing of RT always results in 0ms.  It's only when something gets in the way that it will add to the Response Time.

 

Have you tried another browser?  Try in Chrome.  What are the specs of the machines?  

 

[jackdashack]

Do you think something might be going bad on Verizon end. I remember this sort of thing happened once when a module or node or something was beginning to go bad downtown. At this point it's not happening enough to really notice it. Yes, this last comment of mine, test, was on the Windows 10. Sorry, I am doing business on the win10 and just ran a test. AND, I never ignore the results from testmy.net. But I don't know what it might be saying like you would .... ??

[CA3LE]

Run ping from the command prompt and it will help us understand more.  Paste the results here.

[jackdashack]

This is a ping test from the kitchen XP on the slave router that is connected by direct shielded Ethernet. A short Ethernet cable to the XP. I'll go do the same on the windows 10 at the primary D-Link Router supplied by Verizon

[jackdashack]

This is the same test run on the Windows 10 connected by Ethernet to the primary D-Link router. An frankly Ca3le I should fess and tell you I'm not real savvy on all the technical issues you are speaking of. Somewhat but not all. 

 

[CA3LE]

ping east.testmy.net 

[jackdashack]

Here is the XP on the slave ... the same test on the Windows 10 just next to it. I see a difference but not sure what it all means. You are right, it is not giving the same stats as the 4 other speed tests using flash. I have no idea what this might tell but I suppose to a trained eye some good. About 30ms average difference of the two.
 

[CA3LE]

You've confirmed what TMN is telling you.  The two 16ms readings tell us that.  Your actual ping isn't far off from what TMN Response Time shows.  Run it longer than 4 hops, try ping -c 20 east.testmy.net

 

Just like with your Response Time, your ping is going back and forth.  That's never a good sign.  Room for improvement for sure.

 

Thanks for showing me that the test is working.  

 

So yeah, for all we know... the other tests out there that show 'ping' are showing you the best ping out of a series.  TMN shows you everything that it sees.

 

Hope this helps!

[jackdashack]

24 minutes ago, CA3LE said:

Just like with your Response Time, your ping is going back and forth.  That's never a good sign.  Room for improvement for sure.

 

Thanks for showing me that the test is working.  

 

 

Room for improvement? Gee .. how. I just spent a whole week installing shielded DSL line, dedicated no other phones on it. I wonder what else I could do to "improve" the stats. I never doubted your tests or your manners of doing things. I was just wondering about the differences being shown. I'm still not sure I understand all the input. If brother was here he'd be able to break it down. Alas, he just left a few hours ago. We thought we'd really done something all the work we did this week. We cut new ends on the telephone lines on the block inside. Brightened the posts, shielded dedicated DSL line. Ethernet line from the primary to the slave. Gosh .. there's more? Not to bother, you've take enough time here with me. Thanks much!  jack  ":-Dx

 

Edited December 11, 2015 by CA3LE
fixed quote

[CA3LE]

Could easily be on your ISP's end... it may be totally out of your hands.  No matter how good you make it your end, you're only as strong as the weakest link.

 

It is fluctuating heavily and that's never good.  I would definitely look deeper into that.

 

Aren't you convinced yet that those other sites are a waste of your time?  Why do you keep visiting them?

 

31 minutes ago, jackdashack said:

Not to bother, you've take enough time here with me. Thanks much!  jack  ":-Dx

 

You're never a bother, I appreciate questions, comments and suggestions. Keep 'em coming!

[jackdashack]

I only use the others as a reference. Most times actually Testmy and they are not that far apart. But the one I always watch and trust is Testmy and it's results. Now I don't know what to do. I guess I'm done. Verizon is all we have. No FIOS will ever be here and I will never go Comcast again. 3 other neighbors in my immediate vicinity have the same service. Would love for him to see if they've got these fluctuations. What simple test could I ask them to do to check for just fluctuation? Is satellite internet a bad proposition. It's more expensive and I have to consider that. But these fluctuations ... I can't say I really notice them. My service seems to be steady. ???

[mudmanc4]

May I suggest running a traceroute to east.testmy.net, and post the results here please.

 

tracert east.testmy.net

 

[jackdashack]

1 hour ago, mudmanc4 said:

May I suggest running a traceroute to east.testmy.net, and post the results here please.

 

tracert east.testmy.net

 

The one with the lower ms after the Timed out is the XP on the Slave and the one with the way high ms after the timed out is the Windows 10.  Are those "timed outs" the issues. Does that signify the problem. If so I can get the neighbors to run this tracert too and see if it's a commonality in this little cul de sac.

 

 

Edited December 12, 2015 by mudmanc4
img duplicate

[jackdashack]

Sorry, I duplicated the images. 4 images, 2 of the same each. ":-\\

[mudmanc4]

The nodes that timed out are not always abnormal, it's simply the 'security' setup to not respond.

 

What is more concerning appears to be the XP box, where the 10 box has not this issue.

Yet you are getting the same test results from both machines, correct me?

[jackdashack]

1 minute ago, mudmanc4 said:

The nodes that timed out are not always abnormal, it's simply the 'security' setup to not respond.

 

What is more concerning appears to be the XP box, where the 10 box has not this issue.

Yet you are getting the same test results from both machines, correct me?

 

Actually, I mentioned I duplicated the images and confused you. The XP is the one reporting the ms in the 16ms area after the Timed Out's. The 10 box is the one with over 100ms after the Timed outs. I did remove the slave from the 10 box once and ran the same test without the slave being attached. Same results so my laymen mind said the slave has little to do with it. So knowing the 10 is reporting much higher ms after the timed outs what is it that concerns you. I really have no clue, it's a "no clue" question from me. lol

[mudmanc4]

ok no problem, lets simplify this.

Take one machine out of the discussion momentary.

 

Is there one of th[e] two machines which can be connected to either or internal 'nodes' or switches as you are referring to?

If so lets use that one to test each 'switch or node'

 

[jackdashack]

I'll be gone about an hour. But I'm not quite sure what you mean yet. As I mentioned to CA3LE I have an slightly above the average pc user understanding. You mean turn one pc off,  I'd probably eliminate the XP on the slave router?  That would mean disconnect or pull he Ethernet line from the D-Link Verizon Router which is/was to the XP? As far as nodes or switches I'm clueless. And then when one is completely out of the picture what then? Run the tracert test or .....  Back in an hour. Thanks so much for your time ............

 

[mudmanc4]

Is it possible to get the topology of your network, in other words:

 

Modem---> switch --> windows machine

 

Or Modem --> switch --> switch --> XP windows machine

                           |

                       windows 10 machine

 

It all good, we should have the above before at least myself, has a better understanding of the layout.

 

Also, when referring to the 'node', this could mean any number of network 'devices' in the path, a switch, a router, to keep it simple, it's a 'hop' that must be traveled to reach the destination. Which is what the tracrt shows, each 'hop' from 'node to node' is showing the 'latency' or speed / time to reach there and back divided by 2.